Printable View
Also he answeredQuote:
Originally Posted by Jazzman
So his comparison is not clock by clock.Quote:
You don't believe its 20-30% faster than barcelona. The chip has 30% higher clock, 300% more L3 cache and some IPC improvements over Barcelona.
BTW, Barcelona scales better than Intel Harpertown when you run on all cores.
I wonder if AMD has learned a few things from their years of experience with an integrated memory controller. Also, thinking out loud, I wonder if Nehalem's reported stablity problems relate to their IMC-CPU approach and the "new DDR3 mobo"/"new cpu" learning curve? Not that I have heard this is a showstopper, only a few hiccups. :shrug:Quote:
Originally Posted by Eson
PROOF ?Quote:
Originally Posted by Nedjo
Actually he is wrong since Shanghai doesn't run at 30% higher clock.. It's 2.7Ghz versus 2.5Ghz,top clocks for Shan. and Barc.Difference is mere 8%.It has been already said,numerous times by AMD officials,that in various server workloads top clock Shanghai will outperform top clock Barcelona by 15-35%,which translates into IPC boost between 6 and 25%.They said on average it will be closer to the later figure(25%).Quote:
Originally Posted by justapost
You already could have seen the preliminary pricing at 2 sites.Quote:
Originally Posted by demonkevy666
http://news.softpedia.com/news/AMD-0...rs-96862.shtml
For reference ,here's the top Barcelona 8xxx series price today:Quote:
The new AMD processors have not been officially introduced by the manufacturer, but they are expected to be launched on November 13th, according to industry sources. The chips that got listed at retailers are the Opteron 837X and 838X series, and it seems that they are not cheap at all. The Opteron QC (quad-core) 8384 is listed on two sites, PC Connection and Buy, at $2,509 and $2,240, respectively. The chip is supposed to feature a core speed of 2.7GHz and comes with 75 watts power consumption.
Another new AMD server processor listed is 8385, which has the same core frequency but a faster system bus. PC Connection offers it for $2,509. The site also lists the 8382 (2.6GHz), 8380 (2.5GHz), and 8378 (2.4GHz) CPUs, priced $2,177, $1,768, and $1,360, respectively. The prices will differ for sure from the official AMD pricing. Also listed are chips from the Shanghai Opteron 230X series, namely the 2382 (2.6GHz) and 2380 (2.5GHz), priced at $1,019 and $814 respectively.
Quad-Core
AMD Opteron Model 8360 SE $2,149
Well lets hope Deneb is no where near those prices!
Yeah these are Shanghai parts.
For desktop i'd expect a 250-300$ range for two models-2.8 and 3Ghz.
Of course, the FX-82 Deneb at 4.4 Ghz will cost a little more.... :D SOURCEQuote:
Originally Posted by informal
That Deneb FX part at 4+Ghz is a pipe dream :).Quote:
Originally Posted by Jazzman
On the other hand,the upcoming 3Ghz BE version should be able to hit 4Ghz with good air or WC,so that one will be a kicker :D
Interesting..........very interesting.Quote:
Originally Posted by Jazzman
I feel like alot of people that dont run AMD setups are lurking:p:
You mean like this:Quote:
Originally Posted by informal
http://i37.tinypic.com/309rm8m.jpg
Hmm maybe he meant 3GHz vs 2.5GHz.Quote:
Originally Posted by informal
25% * 3GHz / 2.5GHz = 30% :)
Someone posted on this thread but I can't find it that if the FX line used metal gates in transistor design they could get a significant FX-type boost from phenom architecture. Thereby getting Deneb FXs in the 4+ Ghz range.(Which is my "pipedream".)http://i35.tinypic.com/2z3ngrd.jpgQuote:
According to Intel the new technology provides for 30% less switching power, 20% faster switching speed and ten times less current leakage in comparison to previously used materials.
Nope,there will be *no* server parts that will be running at 3Ghz,not until Q2 at least.He is speaking about server parts,i think this was pretty much clear.The highest speed bin for Shanghai will be SE part in February and it will work at 2.8Ghz.Quote:
Originally Posted by justapost
The official was speaking about top clock available Shanghai part(2.7Ghz) versus top clock Barcelona part we have on the market today(2.5Ghz).
As a complete noob to the opteron series what is the difference between the 23xx series and the 83xx series that results in such a huge price chasm between the two?
If you type in Opteron and then the CPU numbers into google there are a load of places online already taking pre-orders and the prices are inline with what has been stated above. But no one has listed a decent product specification yet so I was wondering what the differences between the models are other than just clockspeed.
2xxx is two sockets only.Quote:
Originally Posted by Chris_redfield
2xxx = 2-way supportQuote:
Originally Posted by Chris_redfield
8xxx = 8-way support
I can't give a good explanation why 8'way parts are that expensive, but I guess AMD has no competition there or something;)
This was my thought, exactly, after hearing the rumors about FX. AMD indicated months ago that the first Denebs will use immersion litho and silicon germanium gates with advanced strain techniques, but chips further down the roadmap may use ultra-low K dialectric interconnects and high-K metal gates........possibly FX?:shrug:Quote:
Originally Posted by Jazzman
http://www.amd.com/us-en/assets/cont..._Press_Q-A.pdf
I'm happy with a 3G BE cpu (hope they'll exist), and if it hits 3.5 I'll be happy already.:up:Quote:
Originally Posted by informal
Ah right so if you wanted to use one of the four socket F motherboards you would have to use the 8xxx series CPU's.
i have a feeling that both the 2.8ghz and the 3.0ghz am2+ cpus are gonna be BEs. if not then that is a huge mistake since these might be the last am2+ quad cores and the first of the new series. amd can't come into this silently they gotta come in with a huge bang and get everyone's attention. which could be why they are not saying anything right now so no one knows what to expect. but we have seen cpuz screens of the cpus and they all have unlocked multipliers.Quote:
Originally Posted by possessed
That's right.:)Quote:
Originally Posted by Chris_redfield
Are these numbers any reason to be concerned? =(
http://www.theinquirer.net/gb/inquir...ghai-benchmark
I sure hope the above isnt a real representation of the shanghai otherwise AMD really needs to get their crap together, and I love AMD and Im saddened...lol someone say its not true!
you can't trust them. besides everyone already knows that barcelona is already doing fine in the server market. they are saying that intel is dominating them there which is not true. i wouldn't believe it at all until i see a credible source. then they go on to say amd is using 8 sticks of ram instead of 4 then going on and saying with 4 the performance would be better. well maybe it would be better but not by much.Quote:
Originally Posted by GenTarkin
I am not worried about that report....And not just cause it is the enquire...Quote:
Originally Posted by GenTarkin
I have a friend who works for a company building servers for large corporations and most of the systems they sell use AMD CPU's.
I think he also said they sell more systems using the RSX cpu than they do Xeons. Though I tend to get a bit confused when he gets going.
Why are you worried? It seems you are not understanding the numbers in that report.AMD Shanghai 2P server working at 2.7Ghz is faster than 3Ghz 2P Harpertown system!What is better on intel side is the power draw since they use the 2x less RAM and Low Voltage edition of QC chips to post those 1000K scores(that still underperform in this particular test).They even use single socket intel quad core systems..Quote:
Originally Posted by GenTarkin
So in pure performance,a 2P Shanghai @2.7Ghz scores 338K,the same as 2P 3Ghz Harpertown.And still manages to have 20% lower power draw.The 1000 scores from intel camp are using less RAM(8GB versus 16 in AMD server) and are configured with a lot slower 45nm intel chips(2.5/2.66Ghz Harpertowns which in this test could be beaten with 2.3Ghz Shanghais).
Just a little something I played with this week...Too bad I can't say what it was, CPU wise anyhow. :)
This has to do with virtualization, right?Quote:
Originally Posted by freeloader
Absolutely...virtualization and a certain AMD CPU. :DQuote:
Originally Posted by m411b
Thats what I like to hear!
Oh man,32 processors and 73Ghz total CPU resources :eek: :D.I envy you :)
4 hosts as in sockets?:confused:
I think it is 4 hosts as in 4 cores...Which would make the chip used in that screeny a 2.3Ghz chip, correct?Quote:
Originally Posted by AMDDeathstar
Its hardware assisted virtualization to create the illusion of physical hardware to achieve the goal of operating system independence from the hardware.
Looks like you run 4 dual socket servers with two 2.3GHz CPU's on each.Quote:
Originally Posted by freeloader
Looks that way.
4 hosts and 32 processors.
Must run super pi and cinebench well. ;) :p:
Took an deeper look on those >1000 spec_power intel results.They had disabled few xeon features whom might not help for spec_ops but power consumption.
AMD had just disabled Hardware Prefetch. Looking at other intel testings those features are allways disabled for benching.Quote:
System Under Test Notes
* Each JVM instance was affinitized to a socket.
* Using the local security settings console, "lock pages in memory" was enabled
for the user running the benchmark.
* C1 Enhanced Mode Enabled in BIOS.
* Hardware Prefetch Disabled in BIOS.
* Adjacent Cache Line Prefetch Disabled in BIOS.
* Intel EIST support Enabled in BIOS.
* Fan speed control Mode 4-pin (server) in BIOS
Edit: Never mind Adjacent Cache Line Prefetch is disabled by default
http://software.intel.com/en-us/arti...ed-prefetchers
This Thread has now been de-escalated to a Elevated Risk
http://i278.photobucket.com/albums/k...d/elevated.jpg
http://www.techspot.com/news/32325-a...his-month.htmlQuote:
Yesterday marked the launch of Intel’s new Core i7 processor series based on the Nehalem microarchitecture. Reviews and benchmarks poured in from around the web, describing Core i7 as extremely powerful and a worthy successor to the Core 2 series, due in large part to the CPU’s on-die memory controller and Intel’s new QPI technology.
AMD of course is not standing still and has recently leaked some more details of its 45nm quad-core competitor on the desktop, Deneb. The company will reportedly launch two such chips this month, the Phenom X4 20550 and 20350, with core frequencies of 3GHz and 2.8GHz respectively.
It’s not yet known how well they’ll perform against Intel’s latest offerings and AMD has already claimed that initially they don’t intend to compete at the enthusiast level, but at least the company is making the chips compatible with current AM2+ motherboards via a bios update – so unlike the Core i7 a motherboard upgrade is not necessary. AMD will also launch six 45nm triple-core CPUs in the first quarter next year, followed by six higher-end 45nm quad-core CPUs and four entry-level processors sometime in the third quarter of 2009.
Related Stories
* AMD Deneb CPUs to start at $300
err mmmm errr yes Chad Boga, please tell me how i can make some friends? maybe in the intel forum? while your there, maybe you can make some too, and :banana::banana::banana::banana: right off? lolQuote:
Originally Posted by Chad Boga
don't feed the troll.Quote:
Originally Posted by soundood
The trolls are the imbeciles determined to turn this forum into a mirror image of the mentally disturbed AMDZone.Quote:
Originally Posted by biohead
Why do we care what you think about this forum?Quote:
Originally Posted by Chad Boga
Oh and if you have some unfinished business with amdzone go there and sort it out,stop mentioning that website in almost every post of yours since it's getting old.
As you can see from the posts in this thread,even intel users came in to tell you to stop your useless trolling.Take their advice and stop boring us with your useless rants here.
My postings other than responding to irrational fanboy attacks made against me, have been to discuss the unlikelyhood of Deneb being 15 to 20% better in IPC.Quote:
Originally Posted by informal
For this great sin, mental midgets such as yourself have referred to my posts as trolling.
If everyone would just ignore this guy, I am positive he will go away. Possibly to AMDZone!
Everyone just ignore his uselessness. If he posts, just ignore it, do not respond. Don't give him the pleasure of a response unless it is to something useful and non-impuning to AMD or another user.
that price is from an a september article. hope that it has come down a bit now. The $250 price would be perfect IMO for the 3GHz BE. (That would hit in the €200 mark, although i might flex it up to €250)Quote:
Originally Posted by Final8ty
done:Quote:
Originally Posted by m411b
This message is hidden because Chad Boga is on your ignore list.
brw, nice mark
Posts: 666
I have an intel rig at the moment but...but sell it in heart beat if deneb delivers the goods...but being a better over clocker than the phenom isnt gona sell me i want the 15% to 20% performance clock for clock...thats being claimed...its all just hype until the chip actually shows its face and gets thrown on the test bench at toms or any ware site that is trust worthy.....
I think techreport.com has very good reviews...they have about every bench test and more real life usage then most...They did a very good i7 review.Quote:
Originally Posted by VOID WARRANTY
yea i will have to check them out...its not that easy to find unbiased reviews :ROTF:Quote:
Originally Posted by charged3800z24
Good point m411b:up:Quote:
Originally Posted by m411b
that's insane to think about.Quote:
Originally Posted by informal
a 2.3ghz Shanghai beating a 2.5ghz harpertown.
kinda odd. why would amd publish benchmarks showing the worst possible situation? but i know for a fact that amd is doing fine in the server market atm so theres no way their scores would be so low. who knows what they are up to.Quote:
Originally Posted by Aussie FX
But I like to read phanboi poasts!:ROTF: I cant be arsed to ignore people though, you ignore someone and then they just go cry over you while you cant read it. Meh:rolleyes:Quote:
Originally Posted by possessed
I say, dont post anymore:p:Quote:
Originally Posted by possessed
I thought about it, but just cant do it to yall! :DQuote:
Originally Posted by Rammsteiner
666 is a good round number though, huh..
where you are moderator?Quote:
Originally Posted by Chad Boga
Chad have you thought of killing yourself and your family?
i donno it seems apt.
how are those voices anyway? still screaming at you? :rofl:
The kind of dribble that only seems to come from AMD fanatics.Quote:
Originally Posted by soundood
How sad for you that you are so socially inept.
Well, you continue posting in a forum where nobody wants to hear your useless, Intel-biased mumbo-jumbo. What does that tell us about YOUR social skills?:DQuote:
Originally Posted by Chad Boga
What you are saying is incorrect on many fronts.Quote:
Originally Posted by knopflerbruce
I reject the notion that this forum must be a mirror image of the idiocy that one sees at AMDZone.
This forum is for discussing AMD related matters, NOT simply drinking the AMD kool aid and only delivering gushing praise for AMD.
My understanding of this shows that I am normal and well balanced, but portrays my detractors in a very poor light, where illogical fanboyism is their stock in trade.
Dude Just STFU and get lost to your intel section :mad: :moon:Quote:
Originally Posted by Chad Boga
Can we drop this yet... for fairness, I feel a lot of people did idg a litle deep intop Brad...I don't beleive he deserved all that he got..now blows have been traded back n' forth, but c'mon, This was a thread about Deneb "just around the corner?"..anyone with general discussion about the Deneb where abouts..feel free to tingle our senses :p:
sorry, I have a 6 yr old, 3 yr old, and a 7mth old going crazy over here..and this board is where I get away from all that.. and LOL...this board remeinds of my kids..
Hmmm... is this something to worry about or:p:Quote:
Originally Posted by charged3800z24
sinse we are off topic.. did anyone see this LOL
'
http://www.fudzilla.com/index.php?op...10358&Itemid=1
"Self-Owned" is the first thing that came to my mind :rofl:Quote:
Originally Posted by charged3800z24
Fud is always good for a laugh :up:
profile page of cb
duh ? :D
Patience is a virtue
it may be one but i don;t got it :up:Quote:
Originally Posted by demonkevy666
OpenBSD founder Theo de Raadt publicly denounced Intel’s Core 2 processors on the OpenBSD mailing list. Raadt cited 38 pages of processor errata from Intel’s published CPU specifications (PDF).Quote:
Originally Posted by Chad Boga
“These processors are buggy as hell, and some of these bugs don’t just cause development/debugging problems, but will *ASSUREDLY* be exploitable from userland code," Raadt said. "Some of these are things that cannot be fixed in running code, and some are things that every operating system will do until about mid-2008, because that is how the MMU has always been managed on all generations of Intel/AMD/whoeverelse hardware."
This is why I will never buy a Core2Duo chip. Dothan is OK, but I don't touch core2duo or core2quad.
More power to you. :ROTF::ROTF::ROTF:Quote:
Originally Posted by mockingbird
In all seriousness....How Old are you??Quote:
Originally Posted by Chad Boga
:confused:
Do you think the post I responded to was a credible one?Quote:
Originally Posted by charged3800z24
most likely not because after reading that post it gives you the impression that core 2 chips are absolutely horrible when they are far from that.Quote:
Originally Posted by Chad Boga
That's my take also. I think the author is pissed because open source systems don't get the respect from cpu manufacturers and/or the mobo manufacturers (i.e. bios updates) that are garnered from the big operating system companies (MS) and big application developers.Quote:
Originally Posted by roofsniper
Also, I think as a result AMD probably tries harder which is why the TLB bug was such a BIG disappointment. My take is that the server business would not tolerate cpu errata that was not addressed by operating systems or mobo manufactures. (One wonders how much "accommodation" AMD gets in this area.) If AMD produced high fault cpus, then Sys Admins would drop AMD in a heartbeat and kill the AMD business model. :bsod:
no, I was just asking....you seem a bit bent on hanging around here and not adding anything usefull...not that everyone here has acted completely mature..but stil.....Quote:
Originally Posted by Chad Boga
So I am supposed to let a gang of morons run me out of town?Quote:
Originally Posted by charged3800z24
Wake up for :banana::banana::banana::banana:'s sakes.
You are not being consistent in respect of other posters by the standards you are judging me on and because of that, I feel no need nor is there a need, for me to justify myself to you.
I shouldn't need to point out that because I had the temerity to think Deneb would only get a 10% IPC boost, that this has attracted an insane level of vitriol being directed at me by various nutjobs and you are falling into the trap of being their useful idiot.
Step out of the Twilight Zone people.
I just started to look through this..on AMD's 45nm
Then why all the snappy post in responce to other people...If you are above these shinanigans then why partake in them....I read through this thread a few times,, (and yes you are aloud to stave your feelings on Deneb's projected performance...), you seem to have less useful info then a statement of why you feel it is going to be so slow.. If people want to dream, let them. I seen a lot of high hopes for i7 in the Intel threads, was that wrong? If someones hopes are defeated once 45nm chips come out from AMD, thet let it be. You can't change anyone's mind.. Besides, all we have are reports and speculation..if we sit and pretend none of it is true, how does that make anything better. No one thought 4800 series GPU's would force nVidia to lower pricing the way they did due to extreme compitition that was never there from the past. Again, why are you so bent on trying to prove something in here??Quote:
Originally Posted by Chad Boga
Hello! This is my first post in this forums. I`m from Argentina.
I have reading almost all the threads about deneb in xtremesystems and I don´t know if someone had already said something about the great hint that AMD has thrown on us with the model numbers. If you you compare this numbers with the model used for Agena you can say that when they launched it they already knew that the fastest part would be the 9950. So I think is safe to say that AMD could and probably would make a 3.4 ghz model. Maybe the FX name is reserved for something faster too:shrug:. I know that there is some info about Deneb running at more than 4Ghz on air, but the models number is a good thing to look at right know. I can not help but think that AMD is doing the same move that did with ATI and the HD48xx series. What do you think about this?
Do you have reading comprehension problems?Quote:
Originally Posted by charged3800z24
There is no equivalency here between my posts and those of my detractors.
I haven't said that if you think Deneb will get the IPC boost AMD are claiming that you are a troll who should be banned and various other insults, so why is that okay for others to say that about me for thinking Deneb won't get as big an IPC boost as AMD are claiming?
Also, what on earth is with your ridiculous argument that me stating my opinion on Deneb is trying to crush other people's hopes for Deneb?
By that "logic", debates and discussions can never take place because you are calling for ONLY uncritical positive praise for AMD.
That is not how a forum like this on a site like this is meant to operate, so stop asking me stupid questions like why I am here.
I am here for the reasons this forum exists in the first place, for discussion and debate.
That a group of fools think this forum should be a mirror image of an AMD propaganda forum like AMDZone, is not something I will adhere to and I suggest once again, that you stop acting like their useful idiot.
:welcome:welcome to xtremesystems. i think amd is reserving the fx for their high end deneb chips. they are most likely not gonna put an fx in the launch and i bet after the launch stabilizes a little and if the company gets into a better financial situation then they will make some high end fx chips to see if they can take the performance crown from intel. right now amd seems to be focusing on everything but the high end.
you seem to be the only one that is a fanboy and refuses to believe the facts. im completely fine if u think that the ipc won't be as high as amd says but u can't say others are wrong because they are just following what amd has said and this time around it looks like amd has gotten back together. its fine if you think that the chip will not be as good as most people think it is or what amd says but what i don't understand is how u can attack people's opinions when they are just going off what they have been told and going off of facts. i think everyone needs to calm down. you need to stop attacking people chad and stop with the name calling and live up to what you have said:Quote:
Originally Posted by Chad Boga
if you notice the only people that are making comments towards you are the ones that you have attacked.Quote:
I am here for the reasons this forum exists in the first place, for discussion and debate.
Chad ,who ever claimed a 25% IPC improvement for desktop parts?? We are talking about Shanghai if you haven't noticed until now.It's a server part and server workloads are very different from desktop.
We have very limited sources of information about Deneb and most people here are just stating their opinions.
So quit beating the dead horse here and go to your beloved amdzone already.I'm sure you would feel as at home there with a lot of space for useless ranting.I don't know about their ban policies but i doubt you would get banned.
Sorry, what exactly are "the facts"?Quote:
Originally Posted by roofsniper
But I can say as I have that I think the 15 to 20% IPC gains will not materalise across the board on any kind of reasonable average.Quote:
but u can't say others are wrong because they are just following what amd has said and this time around it looks like amd has gotten back together.
You have got it completely arse about face, if people stop attacking me first, then I wouldn't have anything to respond to.Quote:
you need to stop attacking people chad
What absolute bullsh1t!!!!Quote:
if you notice the only people that are making comments towards you are the ones that you have attacked.
Do you want to explain the efforts of the following clowns:
keithlm
soundood
informal
Throwed
Oliverda
Eson
Gener_AL (UK)
B.E.E.F.
biohead
knopflerbruce
Blaber
All of them attacked me first.
I have never mentioned 25% in any context, so I can't help you here.Quote:
Originally Posted by informal
So this thread title starts off with "Deneb" just to mislead?Quote:
We are talking about Shanghai if you haven't noticed until now.
As am I, but for some reason I am not allowed to apparently.Quote:
We have very limited sources of information about Deneb and most people here are just stating their opinions.
Yet another nonsensical statement by you.Quote:
So quit beating the dead horse here and go to your beloved amdzone already.
I dont see what anything said here has to do with the release date of Deneb?? The opinion formed here about how bad or how good Deneb does is off topic. The thread title is "Deneb just around the corner?"
So what does Denebs performance have to do with that?
If anyone here wants to talk about how bad Deneb does could you please start a new thread?
No need for this comment,, not on topic..Quote:
Originally Posted by Chad Boga
Again, is this postive at all.. no.. you are not suggesting it will be slower, you must have already know it is or why would the disappointment be his...Quote:
Originally Posted by Chad Boga
They are aloud to this opinion..Some are a little sensitive yes.. but feeling strongly for something is not horrible...you are stongly holding to the "will not be spectacular" notion.. and that is fine.Quote:
Originally Posted by Chad Boga
Do you have the breakdown of both process's 65nm and 45nm to prove that there is nothing that "might" do more then this? And not being to picky but Deneb is Phenom, 65nm is agena and 45 nm is Deneb...Quote:
Originally Posted by Chad Boga
So here you are pressing that AMD can't even match Intel, Even though Intel has same gen processors running sdame clocks but higher cache count and thus faster, more so with the "die shrink"Quote:
Originally Posted by Chad Boga
Here again why dunb yourself down to someone else's level if are not here to attack with comments back like "mental midgets"Quote:
Originally Posted by Chad Boga
Like I said, feel free to speculate, but brings facts to why, besides that "it is just a die shrink" If there is proof that nothing else they have done from one process to the next is not enough to aid in any general boost then lay it out. We can discuss it. There are still soem of us wiling to do this and speculate some more till results hit the net..Quote:
Originally Posted by Chad Boga
I never said that everyone else was not responding wrongly towards you, but you have made some snappy post in regaurds to them. I have some above... Not all your post were bad, and yes there were more in regaurds to you from others here..and I apologize for that. No one should feel unwelcome here cuase they have another view on the matter. But can we get back to the facts and or hear say floting around.. and discuss what might be..and if someone wants to dream.. can they be free to.Quote:
Originally Posted by Chad Boga
No one without direct access to Deneb results knows whether the IPC improvements will come in at the high end or low end of expectations, so I can't see any reason for why one view should be hailed and the other condemned.Quote:
Originally Posted by charged3800z24
As I have already stated, I hope that the people believing the high end improvements come about are right, but I have expressed why I suspect they won't be.
The reason for this is at least the following:
1. I am not convinced that the improvements to the core will be that much more significant than is usually the case when a core gets a shrink and cache added.
Yes AMD have stated that they are going to tidy some things up, but do you really think that doesn't happen with other shrinks?
2. What has been the best IPC improvement we have seen from a combined shrink, cache increase and problem fixes we have seen in the last 15 years?
No one has been able to come up with another time where from what Deneb will be going through, IPC has increased to the extent AMD are claiming.
Shangai is Deneb, just for anyone thinking they are different. The only difference is that Shanghai is validated as a server part and wont use plain old DDR2/3(though thats probably more a bios/board implementation issue).
Shanghai/Deneb is more than just a die shrink, AMD has already stated this before.... However, I get the feeling it is more tweaks rather than any large wholesale changes. Eg, widening data paths, or enlargening TLBs etc.
Some parts they are transferring are not really optmised much. AFAIK there is something about them using the same cells for the L3 cache at 45nm as they did at 65nm, so the L3 cache density won't improve that much. This makes the die larger than it would be. Intel has always had the best cache density unfortunately.
However there are some drawbacks, this is supposedly their 1st implementation of the 45nm so clocks etc wont be as good as they should be, however the yields are good and clocks are better than expected.
I know magny cours the 6die is the 1st to use improved 45nm process, dont know how quickly they will shift the rest of production to that.
Im still putting my money on an avg ipc increase of 10-15%, closer to 15%. However its the better than expected clocks they will get that will help close the gap. Of course there will be selected benchmarks, esp on server Shanghai/Deneb will look really good.
However Nehalem 2P servers come out 2009, so AMD won't have the lead for very long.
Finally, I really like AMD, and only use AMD, but anyone thinking Shanghai/Deneb has a chance of beating Nehalem is smoking crack. Note that in the leaked internal AMD memo even they did NOT claim absolute performance leadership. Just performance per watt leadership.
AMD next chance to take the leadership crown is with Bulldozer, until then they will be competitive on price and performance per watt but thats it.
there are picture some where the die is actually slightly smaller now, in one video he stated improve SSE.Quote:
Originally Posted by EvilBlitz
:rolleyes:Quote:
Originally Posted by Chad Boga
yea they are pretty much the same but talk about them is different. people might say shanghai has a 25% ipc improvement but that would be in server applications. the same 25% ipc improvement can not be compared to the performance of deneb. but if shanghai does well then most likely deneb will do well.Quote:
Originally Posted by EvilBlitz
Chad, I suggest leaving this thread and not returning.
Any more posts beyond my post will results in your removal from AMD section.
demonkey, yes it is smaller, 283 down to 243 mm2. Though at well over 200mm2 it is still on the large side.The 1meg L2 90nm Opteron was large at 230mm2.
Yes I would classify improved SSE as a minor improvement. He could just mean improved support ie new supported SSE versions and or instructions, but not necessarily performance enhanced per clock.
They weren't originally going to have SSE4 till Bulldozer, maybe they changed their mind and brought it in with Shanghai?
roofsniper, shanghai will probably be better for server than desktop I agree, but I still think 25% avg ipc gain for sever apps seems a little too optimistic.
i think amd said at one point it would be 25% on server and it sounds optimistic to me as well but they are designing the chips for servers and the desktops come as an added bonus.Quote:
Originally Posted by EvilBlitz
Found a quote HERE. This may have been quoted in the our Deneb thread but a quick scan by me did not uncover it. This is likely just marketing but I thought it was worth citing for the record.Quote:
Originally Posted by roofsniper
I suspect that Pat's words are misinterpreted by the author. I think Informal's math exercise is closer to how we could see 35% improvement.Quote:
A vice president of Advanced Micro Devices said in an interview that the forthcoming quad-core AMD Opteron processor code-named Shanghai will have roughly 35% performance advantage over predecessor at the same clock-speed. If the information is correct, then AMD’s chips will have all chances to challenges products by arch-rival Intel in the higher-end segment.
Pat Patla, general manager of AMD’s server and workstation chip business, said in an interview with IDG News Service that AMD Opteron “Shanghai” would “give a 35% performance boost over Barcelona on average” and would also trim power consumption by 35%, which automatically transforms into colossal performance per watt improvements over existing quad-core AMD Opteron processors.
Quote:
Originally Posted by informal
a very calm and fair decision, now we need to alert the Mod/owner of AMDzone, that hes on his way.Quote:
Originally Posted by FUGGER
LOL :ROTF:Quote:
Originally Posted by soundood
Anyway its looks to be an exciting 2 months on the way.
Deb 2.8-3.0 + 4GB ram + 4870 would be nothing to laugh at.
Oh man you just made my day :D,that was soo funny :up:Quote:
Originally Posted by soundood
AMDzone is in for a ride :).
Anyhow i searched the spec.org for any newer additions on the SH Opterons and had no luck :(.Maybe the other results will come on the night of the official launch(next Wednesday).
Edit: found an interesting article on the project Ridgeback design cycle.Ridgeback is internal AMD code name for Shanghai
http://www.statesman.com/business/co...0shanghai.html
Quote:
ADVANCED MICRO DEVICES INC.
AMD seeks redemption - and sales - with Shanghai
Chipmaker used lessons of Barcelona in revamping design process
By Kirk Ladendorf
AMERICAN-STATESMAN STAFF
Monday, November 10, 2008
To the outside world, Advanced Micro Devices Inc. calls its latest chip Shanghai. To the engineering team, the usual name for the project is Ridgeback. But the real name for this chip could be "Redemption."
AMD wants to restore some of the luster to its image that was lost when its last flagship chip, Barcelona, was months late and plagued by embarrassing bugs.
The chipmaker was counting on Barcelona to revive its sagging fortunes in its perennial battle with Intel Corp., the kingpin of the microprocessor market.
But Barcelona's public problems had the opposite effect. They cost AMD heavily in lost revenue, lost profits, lost morale and lost confidence from its customers.
Now comes Shanghai and the hope for a recovery of reputation and, maybe, sales.
The new chip began shipping to computer makers in October and will become the brain of new servers that show up this month. The formal unveiling is Thursday.
Shanghai is similar to Barcelona, but there have been dozens of enhancements, including the addition of larger on-board memory caches to spur performance.
AMD says the chip will quickly become the standard setter for Windows-compatible server chips, although Intel will counter with its own high-performance chips in the weeks ahead.
"I am delighted with the chip," said Jeff VerHeul , AMD's vice president of silicon design. He cited the reasons: "A, it is early; B, it has greater-than-expected performance; and, most importantly, our customers like it."
He backs the assertion of CEO Dirk Meyer , who called the new chip "the best server product on the market."
The company has yet to deliver precise performance metrics, but VerHeul said Shanghai will deliver roughly 20 percent more performance than Barcelona while running at the same speed. Plus, it will run faster.
The new chip, like Barcelona, contains four processing engines, or cores, but it is far more complex because it is made with a more advanced manufacturing process.
Analysts are not quite as enthusiastic as AMD.
"They need to rebuild their reputation and re-establish credibility," said Nathan Brookwood with Insight 64. "They need to demonstrate that they are executing and delivering on their promises. This should help there."
AMD is aiming for a financial rebound as well after two years of losses. The chipmaker has sold some nonproductive operations and announced plans to spin off its German factory operations into a joint venture it will own with a sovereign investment company based in Abu Dhabi.
In addition, the company announced last week that it had cut 500 jobs worldwide, including 154 in Austin, as part of an effort to cut operating costs as it aims for consistent profitability.
The problems with Barcelona forced AMD's leadership to take an intense look at its design process and fix the flaws it found. Part of that fix involved a new procedure for designing complex, high-performance chips.
In the past, the company relied mainly on one of its two major design centers in Austin or Sunnyvale, Calif. But the work on the new chip would be split among at least five design centers around the world: Austin; Sunnyvale; Fort Collins, Colo.; Dresden, Germany; and Bangalore, India.
The first four centers would all work on key sections of the chip, and the design center in Bangalore, created in 2005, would be responsible for putting the pieces together in a complete design.
The project was directed by a veteran AMD engineering manager Raghuram Tupuri, who moved to India in 2005 to get the Bangalore design center started. He returned to Austin last year.
The approach was the brainchild of VerHeul, who dubbed it Centers of Excellence. He had tried a similar approach at IBM Corp. on engineering projects involving the design of servers.
Splitting up a major chip design among several design centers can result in faster completion of the design, but VerHeul and Tupuri both say the approach requires extensive communications and personal rapport among the key players in the various far-flung design centers.
"We had people from the various sites spend time together so they got to know each other," VerHeul said. "Developing personal relationships is very, very key."
Tupuri said the global design effort came together as one big team, with engineers eventually feeling comfortable enough to call one another on important questions before, during and well after normal work hours.
"We are standing on the shoulders of the Barcelona team and getting the benefit of the lessons we learned from Barcelona," Tupuri said. The engineers "wanted to succeed and put Barcelona behind them. They looked at Ridgeback definitely as a redemption."
The key objective of the team was to check and recheck the design to improve the odds that the first silicon prototypes of the new design would function well. Achieving that goal saved months in the development cycle.
"We were all keen on getting the first silicon healthy," Tupuri said. "If the first silicon is healthy, then the glide-path (to product launch) is set. Even if we delayed the tape-out (formal completion of the design), we wanted to get the first silicon right."
If Shanghai is as good as advertised, then it should help boost AMD's sales of high-performance chips in the months ahead.
But analyst Ashok Kumar with Collins Stewart LLC says the impact of the new chip might be muted by the weakening world economy, which has put a big dent in sales of servers to major corporate buyers.
"Demand has fallen off a cliff," Kumar said. Shanghai "will let AMD hold its own in the current environment, which is the best they can do."
LOL, laughed my arse off:rofl::clap:Quote:
Originally Posted by soundood
Im really curious what Deneb will bring. Im about to order a Core i7 (yes I know:p:). AMD is fine, but Im slightly done with the low quality crap motherboards. It ain't the first time since C2D/C2Q that a motherboard here went crappy and took the CPU right with it. But well, if Deneb is good Ill certainly buy it anyway, although it wont be in my primary PC though. But hey, in the end it's about the overclocking fun anyway:up:
This is my post from a Shanghai thread in News section:
I found this info interesting:
Courtesy of Johan De Gelas :rolleyes:Quote:
That reassured us that our transaction log disk was not a bottleneck. On a "normal" SLES 10 SP2 we achieved 1400 tr/s on a quad core (an anonymous CPU for now ;-). But Anand's article really got us curious and we replaced our mighty Cheetah disk with the Intel x25-M SSD (80 GB). All of a sudden we achieved 1900 tr/s! No less than 35% more transactions, just by replacing the disk that holds the log with the fastest SSD of the moment. That is pretty amazing if you consider that there is no indication whatsoever that we were bottlenecked by our log disk.
Wonder what that anonymous CPU is :p:
Here are results to compare from his previous article:
http://www.xtremesystems.org/forums/...chmentid=88814
http://www.xtremesystems.org/forums/...chmentid=88812
Does anybody know the overall speed increase of C2D over Barcelona clock for clock? I know I am making a very rough estimate, but I'd like to put some kind of numbers in my head.Quote:
The company has yet to deliver precise performance metrics, but VerHeul said Shanghai will deliver roughly 20 percent more performance than Barcelona while running at the same speed. Plus, it will run faster.